Final upldates for 2.1.3 release.

README

@@ -1,15 +1,36 @@
+Introduction
+------------
+
 gmpy2 is an optimized, C-coded Python extension module that supports fast
 multiple-precision arithmetic.  gmpy2 is based on the original gmpy module.
 gmpy2 adds support for correctly rounded multiple-precision real arithmetic
 (using the MPFR library) and complex arithmetic (using the MPC library).
 
+Version 2.1.x Status
+--------------------
+
 gmpy2 2.1 was extensively refactored. Some of the significant changes are:
 
   * Support for thread-safe contexts and context methods
   * Interoperability with Cython extensions
-  * mpz and mpq operation can release the GIL (controlled by the conntext)
+  * mpz and mpq operation can release the GIL
+    * The current implementation is experimental
   * Improved argument processing
 
+The gmpy2 2.1 series will be the last to offer compatibility with Python 2.7.
+Release 2.1.3 is the last planned release of the 2.1 series.
+
+Version 2.2 Plans
+-----------------
+
+Version 2.2 will drop support for Python 2.7 and older Python 3.x versions.
+
+The primary development focus will be on functions that operate on lists and
+release the GIL. See powmod_base_list and powmod_exp_list as examples.
+
+Availability
+------------
+
 gmpy2 is available at https://pypi.python.org/pypi/gmpy2/
 
 Documentation is available at https://gmpy2.readthedocs.io/en/latest/

demo/parallel_powmod.py

@@ -2,7 +2,7 @@
 import concurrent.futures
 import time
 
-# Test parallel computation of gmpy2.powmod()
+# Test parallel computation of gmpy2.powmod_base_list()
 
 # The demo code is not (yet) optimal since it doesn't take advantage of
 # read-only access to data list that is shared across all threads.
@@ -13,7 +13,6 @@ def create_tests(num_items = 100000, bits = 1000):
     e = gmpy2.mpz_urandomb(rand, bits)
     m = gmpy2.mpz_urandomb(rand, bits)
     big_list = [gmpy2.mpz_urandomb(rand, bits) for _ in range(num_items)]
-    e = 65537
     return big_list, e, m
 
 # Partition big_list into a new list containing a number of sub-list with
@@ -53,7 +52,7 @@ def powmod_vector_nogil(index, vector, e, m):
 
 # Run threaded versions.
 def run_test(function, big_list, e, m, threads, release_gil = True):
-    over_split = 4
+    over_split = 8
     size = len(big_list) // (over_split * threads) + bool(len(big_list) % (over_split * threads))
     split_list = partition_list(big_list, size)
     total_thread_time = 0
@@ -69,13 +68,13 @@ def run_test(function, big_list, e, m, threads, release_gil = True):
     return time.time() - walltime_start, total_thread_time
 
 if __name__ == "__main__":
-    big_list, e, m = create_tests(100000, 32000)
+    big_list, e, m = create_tests(100000, 1024)
 
     # Run a baseline test with releasing the GIL.
     # print("Baseline test without releasing the GIL. ", powmod_list_gil(big_list, e, m)[0])
 
     # Specify the number of threads to use.
-    test_threads = [1,2,4,8,15,16,24,31,32]
+    test_threads = [1,2,4,8]
     # The following demo code is disabled by default. It high-lights the inefficiency of 
     # releasing the GIL for functions that execute fairly quicky
     for i in range(0):